Skip to Content
logologo
AI Incident Database
Open TwitterOpen RSS FeedOpen FacebookOpen LinkedInOpen GitHub
Open Menu
発見する
投稿する
  • ようこそAIIDへ
  • インシデントを発見
  • 空間ビュー
  • テーブル表示
  • リスト表示
  • 組織
  • 分類法
  • インシデントレポートを投稿
  • 投稿ランキング
  • ブログ
  • AIニュースダイジェスト
  • リスクチェックリスト
  • おまかせ表示
  • サインアップ
閉じる
発見する
投稿する
  • ようこそAIIDへ
  • インシデントを発見
  • 空間ビュー
  • テーブル表示
  • リスト表示
  • 組織
  • 分類法
  • インシデントレポートを投稿
  • 投稿ランキング
  • ブログ
  • AIニュースダイジェスト
  • リスクチェックリスト
  • おまかせ表示
  • サインアップ
閉じる

レポート 408

関連インシデント

インシデント 2830 Report
2010 Market Flash Crash

Loading...
Navinder Sarao: The man accused of causing the US market to crash
bbc.com · 2016

Image copyright Getty Images Image caption Navinder Sarao, the so-called "flash crash" day trader has been fighting attempts to extradite him to the US

Navinder Sarao, who traded from his parents' home in Hounslow, west London, has been accused of market manipulation that caused a 1,000-point fall on the US Dow Jones index in 2010.

He faces 22 charges in the US, including fraud charges, all of which he denies - which include "spoofing" - the practice of buying or selling with the intent to cancel the transaction before execution.

Set aside for a moment Navinder Sarao's guilt or innocence. What he did was astonishing.

From a computer in the bedroom he'd grown up in, at his parents' semi-detached house under the Heathrow flight path, this casually dressed 37-year old traded remotely on an exchange in Chicago he had never visited.

In less than five years he made more than £30m.

Computerised trading

Pause for thought. He was operating on the Chicago Mercantile Exchange, a highly computerised market where normal human beings stand little chance.

Computers operating so-called high frequency trading programmes pick up any sign of prices moving up or down - and jump in ahead of normal, slow-motion human traders.

Image copyright Getty Images Image caption Computers will always beat humans when it comes to trading

You, the human trader, may see an order to buy pop up on your screen half a second after it's published on the exchange. It may be big enough to move the market up - so you could make money if you could buy quickly.

But because you're up against computers, by the time you click 'buy' it's already too late. Your brain took half a second to process the visual information - whereas a computer picked it up in a few milliseconds.

The high frequency traders bought, the price rose - and you lost your chance to make money.

In Chicago, the pressure is fierce to see buy and sell orders just a few milliseconds ahead of the other guy.

Millisecond advantage

As outlined in Michael Lewis's brilliant expose Flash Boys, a few milliseconds' head-start is worth so much money that the most determined traders buy expensive property close to the exchange and spend large sums installing super-fast cables - so the signal arrives just a few milliseconds faster.

Why? Because if your computerised trader gets the information two or three milliseconds before others, you can make money.

Image copyright Getty Images Image caption But computerised trading does have a weakness, especially if the computers are programmed to react in similar ways

If it's a big buy order your computerised trader can buy ahead of others, "riding the trend" and adding to upward pressure on prices.

The others who get the information a flash later can't "get in at the ground floor". Instead they buy after prices have already risen. And they're buying at those inflated prices from you - as you sell out for a quick buck.

'Spoofing' the computers

What Navinder Sarao seems to have worked out, like others before him, was that the high frequency trading made the market twitchy.

Because many of the high frequency traders were programmed with similar software - designed, for example, to spot a large volume of buy orders and then ride the trend - they would all do more or less the same thing.

Like a pack of sheep sent one way or another by a sheepdog, they could be spooked.

Or rather, "spoofed". According to the FBI's investigators, Nav Sarao had modified a commercially available piece of software - a trading algorithm, to do just that.

Image copyright Getty Images Image caption The FBI didn't only accuse Mr Sarao of placing fake orders to manipulate the market, it also accused him of helping to cause the flash crash

Using that software and at high speed, they alleged, he was placing big sell orders, sometimes amounting to hundreds of millions of dollars a day - trades with which he never intended to go through.

Instead they were designed to make the rest of the market (the computerised traders) believe the sell orders outweighed the buy orders, indicating the market was about to head down.

Sheep-like, the computerised traders would sell in order to avoid losing too much money as prices dropped. The weight of selling would push prices down.

Meanwhile (according to the FBI) Mr Sarao would have placed a real buy order waiting for the market to drop. He would then buy at the lower price and cancel the sell orders.

When the market realised the sell orders had gone away, prices would bounce back. And Mr Sarao could then sell at that higher price and bank a profit.

That profit might be tiny. But if this operation were repeated hundreds of times per hour, the profits could be handsome.

Crucially, according to the FBI, those orders were fake, designed to "spoof" the market. Sarao didn't intend to go through with them. And that spoofing is an offence under US law.

The FBI didn't only accuse Sarao of placing fake buy and sell orders to manipulate the market. It also accused him of

情報源を読む

リサーチ

  • “AIインシデント”の定義
  • “AIインシデントレスポンス”の定義
  • データベースのロードマップ
  • 関連研究
  • 全データベースのダウンロード

プロジェクトとコミュニティ

  • AIIDについて
  • コンタクトとフォロー
  • アプリと要約
  • エディタのためのガイド

インシデント

  • 全インシデントの一覧
  • フラグの立ったインシデント
  • 登録待ち一覧
  • クラスごとの表示
  • 分類法

2024 - AI Incident Database

  • 利用規約
  • プライバシーポリシー
  • Open twitterOpen githubOpen rssOpen facebookOpen linkedin
  • e1b50cd